I have now reviewed around 7000 posts. Since I review mostly first / late answers (because I see the questions at the front page), you can more or less take this as a statistical sample of what really happens.
- 6960 reviewed
- 1032 edited (and they really needed to be edited)
- 387 were flagged (that's around 5.5%)
I don't think this is a lot. Of this 5.5%, the majority are spam posts and "same problem" answers, which didn't even need to be a comment in the first place.
Do we allow all users to comment everywhere?
No, massive spamming will prevail. New users won't know how to use comments and mistake them for some kind of live chat. Random strangers will ping you asking for help, et cetera. Since comments don't bump posts, they're much harder to review too. This is really not a good idea.
Hold answers for review if the user has less than 50 reputation?
If roughly 95% of all answers are a genuine attempt to solve the problem stated, then I'd say this is more counterproductive than useful.
Impose an answer length requirement for such users, since comments are usually short?
I've flagged non-answers (that really could have been comments) that were almost a thousand characters long. But just not an answer.
Short posts usually land in the "low quality" review queue anyway, and some of them get an automatic flag by the system. This makes it even easier to peer-review them, and moderators can act on them right away (by deleting or converting into a comment).
TL;DR: I currently do not see a problem with the whole "new users can't comment" thing.
If you're really desperate to say something, you'll try to get 50 reputation and then do it. If you manage to phrase an additional answer on that post where you wanted to comment, even better. Or, you could just improve the existing answers – that's the whole point of anonymous or peer-reviewed editing.